Questions on Multilayer Perceptron or MLP

Questions on Multilayer Perceptron or MLP

"Multi layer Perceptron" (MLP) is a concept of Neural networks.

Its basically an input/output model that uses mapping methods for setting inputs data onto a specific set of outputs. Its basically a  'Feedforward artificial neural network'. Here, Feedforward basically means that data is flowing  in unidirectional  i.e in one direction from the  input towards output layer

A multi layer perceptron uses multiple nodes on layers mapped in a directed graph as such each layer is fully connected to the later one. Each node (except the input nodes) is a processing element known as neuron.

Your next question is how to train Multi layer perceptions.

This can be done by various training algorithms. The main

goal for training multi layer perceptron is to find the lowest point in the multi dimensional surface. The concept of error surface also helps in training MLPs, but its very complex. There are many characteristics such as local minima,saddle-points ,flat spots which are to be calculated. While performing it, the algorithm stops at the lowest point, which results in local minima.